Car Damage Classification Flow for Insurance

In alignment with trends observed across various sectors, the automation of tasks through Artificial Intelligence (AI) has gained significant traction within the insurance industry. In the insurance industry, AI is generally utilized in document processing, anomaly detection and vehicle damage evaluation. Damage detection within the insurance domain is typically conducted manually, rendering the process both timeconsuming and costly. The process consists of classifying the records depending on the severity of the damage. Low-cost damage claims are generally processed without the involvement of an expert, whereas high-cost damage claims are subject to expert evaluation. Accurate classification of damage severity is crucial in determining the associated costs. In this study, a machine learning approach is proposed to classify damage severity. The proposed method consists of four phases. In the first phase the input images are filtered and unrelated images are discarded. Then, car images are cropped based on the detected car region. Finally, car images are grouped based on angle and the damage severity is classified. Evaluations conducted on our internal dataset show that the proposed method achieves 88% F1 score. The proposed method offers a solution for automating the vehicle damage assessment process. The method aims to replace manual vehicle damage assessment solutions, thereby reducing costs and minimizing time loss